1 paper · 1 filter
Siddhant Panpatil, Arth Singh, Mijin Koo +3
Vision-language models (VLMs) are now proposed as runtime safety guards for embodied agents in homes and factories. A deployable guard must catch genuinely unsafe situations while…